CN113676357B - Decision method for edge data processing in power internet of things and application thereof - Google Patents
Decision method for edge data processing in power internet of things and application thereof Download PDFInfo
- Publication number
- CN113676357B CN113676357B CN202111002096.1A CN202111002096A CN113676357B CN 113676357 B CN113676357 B CN 113676357B CN 202111002096 A CN202111002096 A CN 202111002096A CN 113676357 B CN113676357 B CN 113676357B
- Authority
- CN
- China
- Prior art keywords
- service
- data
- edge server
- delay
- processing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0803—Configuration setting
- H04L41/0823—Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability
- H04L41/083—Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability for increasing network speed
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Y—INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]
- G16Y10/00—Economic sectors
- G16Y10/35—Utilities, e.g. electricity, gas or water
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/142—Network analysis or design using statistical or mathematical methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/50—Queue scheduling
- H04L47/62—Queue scheduling characterised by scheduling criteria
- H04L47/625—Queue scheduling characterised by scheduling criteria for service slots or service orders
- H04L47/6275—Queue scheduling characterised by scheduling criteria for service slots or service orders based on priority
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
Abstract
A decision-making method and its application for edge data processing in the power Internet of things, including: setting the business importance levels of edge server access: let the set of power services carried by the power communication network be M, where the importance level of the power service m∈M is divided into 4 levels according to the delay sensitivity, and the importance level of the business m is denoted as km, (km≤K and K＝4); building a business processing depth model according to business attributes and network topology: use the edge server c to select the business data volume Sc for data processing, the binary variable 
data basic compression rate Δ and the information attribute set A to build a business processing depth model; solving the optimal data processing decision-making scheme, and set the processing decisions of each edge server for each business.
Description
Technical Field
The invention belongs to the field of electric power internet of things edge data processing, and particularly relates to a differentiated service data flow processing decision method based on edge calculation and application thereof.
Background
With the continuous development of the power internet of things, the service types and the data scale in the power internet of things are also expanded, more and more sensor terminals are deployed in the sensing layer of the power internet of things, and sensing data are uploaded to the cloud platform. However, the data collected and uploaded by the sensor terminal are easily subject to redundancy, errors, loss and other problems due to low energy consumption, insufficient processing capacity and the like of the sensor terminal.
In response to this situation, in recent years, a number of documents have proposed to overcome the above-described problem by preprocessing data before uploading it using an edge calculation technique. Their basic idea can be summarized as the enhancement of grid-bottom processing capacity by introducing edge servers or edge gateways. Ronhong et al published an intelligent measurement terminal data processing method based on edge computing and MapReduce in Intelligent electric journal 2020, volume 48, and proposed a method for improving management capability of data blocks of the Internet of things of electric power by using edge nodes. Chen research and published on volume six of IEEE Transactions on Cloud Computing in 2018, article MapReduce scheduling for offline-constrained jobs in heterologous Computing systems, in an attempt to improve data mining and processing capabilities of the edge layer of the power grid using edge servers. Bai Yi Yang et al, published on volume 46 of the journal 2020, a text of cloud-edge intelligence, an edge computing method for power system operation control and the current application and prospect thereof, and proposed a new mode for improving the fast scheduling and reliable operation capability of a power system by using edge computing. After the introduction of the edge computing technology, the power system can be regarded as that the computing and processing capacity of the system main station is moved down to the edge layer of the network. In the power internet of things, the data processing process of the edge computing mode is beneficial to reducing the data volume of the uploaded data, so that the communication pressure of the network is reduced.
However, although the above work improves the data calculation efficiency of the power communication service, a new problem is brought when the data is preprocessed by using the edge calculation. Compared with the cloud center, the computing power of the edge computing server is very limited, which may cause that a large processing delay overhead is required when the edge computing server performs data processing, and the size of the processing delay is generally in positive correlation with the size of the data amount processed by the edge computing server. Moreover, for some delay-sensitive services, if the raw data is preprocessed, the processing delay caused by the preprocessing may affect the execution of the service.
In addition, chinese patent application, application No. CN2021100901205, publication No.: CN112764835A discloses an electric power thing networking sensing equipment configuration micro-service system and method based on edge calculation, including: the system comprises an edge internet of things agent architecture, a cloud service center and terminal equipment, wherein the edge internet of things agent architecture processes terminal equipment data and performs computing task cooperative processing between the cloud service center and the terminal equipment according to the priority of the terminal equipment data; the edge internet of things agent architecture comprises edge internet of things agent hardware and edge internet of things agent software, wherein the edge internet of things agent software is a micro-service module architecture which is built on the basis of an EdgeX fountain edge computing open source platform by using a Linux operating system and a Docker container engine; the core micro-service module comprises a strategy control micro-service module, a strategy execution micro-service module, a data maintenance micro-service module and a command operation micro-service module. Application No. CN2021100079494, publication No.: CN112835691A discloses an edge data processing method and system for Internet of things communication, which comprises the following steps: acquiring computing parameters of each edge computing device through the cloud computing center, wherein the computing parameters comprise computing capacity and processing time delay; acquiring the request task from the intelligent terminal through the cloud computing center, wherein the request task comprises data computing amount, time requirement, emergency degree and corresponding physical addresses of the collectors, and the emergency degree comprises emergency, urgent and general; and analyzing the request task and the computing parameters through the cloud computing center to obtain the edge computing equipment which is most matched with the request task. However, the foregoing prior art cannot solve the problems that the delay in the edge service processing process is reduced and the efficiency and performance of the network are improved when the power internet of things implements part of services.
Disclosure of Invention
In order to realize the balance of time delay effect of part of services and reduction of transmission pressure of the power communication network, the basic idea of the invention is that the edge server selectively processes data of part of services, thereby reducing the transmission pressure of the power internet of things on the premise of optimizing the time delay effect of all services as far as possible. In order to meet the requirements, the invention provides a decision method for processing edge data in an electric power internet of things, which reduces the transmission pressure of a communication network, optimizes the network performance and improves the communication quality on the premise of optimizing the time delay of all services as far as possible, and the technical scheme is as follows:
the differentiated service data flow processing decision method based on edge calculation is characterized by comprising the following steps:
s10, setting each service importance level accessed by the edge server: the method comprises the steps that the set of electric power services borne by an electric power communication network under the application scene of the method is set to be M, the importance level of the electric power service M belonging to the M is divided into 4 levels according to the time delay sensitivity degree, and the importance level of the service M is recorded as k m ，(k m K is not more than K and K is 4);
s20, constructing a service processing depth model according to the service attributes and the network topology: selecting service data volume S for data processing by using edge server c c Binary variables
Constructing a service processing depth model by using a data base compression ratio delta and an information attribute set A;
and S30, solving the optimal data processing decision scheme, and setting the processing decision of each edge server for each service.
The invention also discloses a differentiated service data flow processing decision method based on edge calculation, which is applied to the power communication network.
The beneficial effects of the invention are: the invention discloses a differentiated service data flow processing decision method based on edge calculation in an electric power Internet of things. The change situation of the data volume processed by the edge server is quantified through an edge Internet of things proxy data processing model, and then the service experiment model is used for analyzing the time delay influence of data processing on the service. This changes the traffic weighted delay and minimization problem into a binary decision problem for the edge server to selectively handle traffic. This problem can be solved by using intelligent optimization algorithms. The invention can reduce the time delay in the edge service processing process and improve the efficiency and the performance of the network.
Drawings
Fig. 1 is a flowchart of a differentiated service data traffic processing decision method based on edge calculation according to the present invention;
FIG. 2 is a flowchart of a processing decision for solving optimal data by applying an intelligent optimization algorithm according to the present invention;
fig. 3 is a performance comparison diagram of a differentiated service data traffic processing decision method based on edge calculation according to the present invention.
Detailed Description
The present invention is explained in further detail below with reference to the drawings and the specific embodiments, but it should be understood that the scope of the present invention is not limited to the specific embodiments.
Fig. 1 is a schematic flow chart of a differentiated service data traffic processing decision method based on edge computing according to the present invention. The method specifically comprises the following steps:
and S10, setting each service importance level accessed by the edge server. The set of power services carried by a power communication network of an input algorithm is set as M, wherein the importance level of the power service M belonging to the M (hereinafter referred to as the service for short) is divided into 4 levels according to the time delay sensitivity (the level is classified and is more than s, hundred ms and ten ms), and the importance level of the service M is recorded as k m ，(k m K is not more than K and K is 4);
the dividing of the service importance level in S10 includes the following steps:
in the power internet of things, each terminal may be applied to different services, and service data collected by the terminal may be uploaded to an edge server accessed by the terminal in a burst or at a fixed period. The edge server is responsible for aggregating and uploading data from the terminal, and determining whether to process the data according to specific time delay conditions.
The communication network topology of the power internet of things is characterized by using an undirected graph N ═ U, C, R and E. Wherein, U is a terminal set, C is an edge server set, R is a cloud routing set, and E is a network link set. Collection of
And characterizing the terminal U belonging to the access edge server C belonging to the C, wherein the number of the terminals in the set is UM c . Edge server c exploits computing power β c For data processing of terminal data information within a coverage area, wherein the computing power beta c Expressed in GHz by the CPU operating frequency of the computer. Since the data volume uploaded by the terminal at a time is usually not large (for example, the data volume uploaded by the pole switch remote signaling sensor at a time is only a few bytes), the process of uploading data to the edge server by the terminal can be regarded as being completed instantly. Edge server c and cloud routing r accessed by edge server c 1 Inter-aggregation link
And cloud routing r 1 And r 2 Inter-forwarding link
Respectively of
And
the set of the power service carried by the power communication network is M, wherein the invention enables the power service M to be in the scope of M (which is hereinafter referred to as M for short)Service) is divided into 4 levels according to the delay sensitivity (classified and above, s level, hundred ms level, ten ms level), wherein the importance level of the service m is marked as k m ，(k m K ≤ and K ═ 4).
And S20, constructing a service processing depth model according to the service attributes and the network topology. Selecting service data volume S for data processing by using edge server c c Binary variable
And (3) constructing a service processing depth model shown in the formula (3) by using the data base compression ratio delta and the information attribute set A. Constructing a service processing depth model according to the service attributes and the network topology:
using binary variables respectively
And an information attribute set a to characterize all information attributes (of | a | types) included in the service M and | M | types of services served by the terminal u. Therefore, the information attributes owned by the service m and the values of the information attributes can be respectively represented by the binary information attribute set A m ＝(a 1 ,..,a i ,..a |A| ) And a set of attribute values B m ＝(b 1 ,..,b i ,..b |A| ) To characterize. In which the binary variable a i Data characterizing a service m includes an attribute a of 1 i And the value of the information attribute is b i 。
For each terminal u, the data information it transmits may be characterized as a row vector
Since there may be multiple terminals serving the same service in the edge server, the data information of the service m in the edge server c may be used as a matrix
To indicate. Matrix array
Has the dimension of
Wherein
Is the number of terminals serving the service m within the coverage of the edge server c. Similarly, the matrix DM can be used when the edge server c receives the data information of all services c To be expressed and matrix DM c Has a dimension of
For any service m on the edge server c, the time delay comprises two parts, namely processing time delay and forwarding time delay. Due to the computing power β of the edge server c c There is a limit, so all the traffic on the edge server c that selects to process data needs to wait for a processing delay
To avoid processing delay
If the size is too large, the edge server c can choose to perform data processing only on the data of part of the services. Processing latency of edge server c
The definition formula is:
wherein, delta c And
respectively edge server c processing delay
Proportional coefficient and basic processing delay, andand the two parameters are both the inherent attribute of the edge server and the received data size S c Is irrelevant. Basic processing delay
It is a necessary delay overhead for the edge server c to initiate the data processing process, and the delay is determined by the hardware configuration of the edge server. Since the edge server c can select to process data only for part of the services, the edge server c selects the service data volume S for data processing c Can be characterized as a data matrix DM c I.e.:
wherein, binary variables
Is a decision variable for the edge server c to perform data processing on the service m. 
And the characterization edge server c processes the data of the service m.
The invention refers to the ratio of the data volume before and after data processing as the data processing depth phi of the edge server c c . However due to the set of information attributes a of the different services m m Composed of different information attributes, and therefore the data processing of different services on the edge server c
As well as different. The invention defines the data processing depth phi of the edge server c c Comprises the following steps:
where Δ is the data base compression rate, related to the effect of the software function itself used to process the service.
And S30, solving an optimal data processing decision scheme by applying an intelligent optimization algorithm, and setting processing decisions (processing or direct uploading without processing) of each edge server to each service.
Because the edge server c may only process data of part of the service m, the edge server c will successively establish two data packets
And
to load the processed and unprocessed traffic data, respectively. Wherein the service data is not processed
Without waiting for processing delay
But can be directly encapsulated in a data packet
And sends the data to a system main station through the power Internet of things
Otherwise, the service data processed by the data processing
It will wait for the processing delay 
And then the packet can be packed and sent to the system master station. According to the processing decision of the edge server c on each service m, the data volume of two data packets of the edge server c
And
respectively as follows:
wherein
And
and respectively loading the processed service data and the unprocessed service data.
After the edge server c completes the encapsulation of the two data packets, the data packet obtained according to the specific routing rule is forwarded to the system master station d 0 The routing path of (1). Two data packets of the same edge server c are agreed in the invention
And
have the same forwarding routes. At the same time, if a certain section of link of network layer
Is simultaneously selected as the forwarding path of a plurality of edge servers c, then the link
Divide its bandwidth amount equally
To each edge server c. In order to characterize the routing relation of the data packet of the edge server, the invention uses binary variable
To characterize whether the data packet of the edge server c passes through the link
The forwarding delay of the two data packets
And
are respectively defined as:
wherein the content of the first and second substances,
the data packet shares the link with the edge server when being forwarded in the network layer
The number of edge servers.
In summary, for the service m, the delay consists of two parts, namely forwarding delay and processing delay. Therefore, the service delay is defined as:
The optimization goal of the invention is to minimize the weighted delay sum of all services, where the delay weight coefficient of a service is the importance level k of each service m . In equation (8), the delay size of the traffic m on the edge server c
Depending on whether the data processing is done or not. Therefore, the objective function of the service delay optimization problem belongs to an integer programming problem, as shown in formula (9):
equation (9) shows whether the edge server c performs data processing on the service m or not under the condition that the objective of time delay weighted sum minimization is considered
The binary decision problem of (2), the corresponding constraints of the binary integer program are:
wherein constraints (10) - (12) are binary variable constraints of a binary integer program. Wherein, binary variables
And if the terminal u serves the service m, selecting the service m, wherein the value is 1, and otherwise, selecting the service m is 0. Binary variable
To characterize whether the data packet of the edge agent c passes through the link
If passing through
The value is 1, otherwise 0.
The intelligent optimization algorithm used for solving the service delay optimization problem is a spider monkey algorithm, and the steps are as follows:
an initial population of N individuals is first generated, where N is determined based on the number of incoming power services. Wherein each individual is a C-dimensional vector, and the ith individual is recorded as SM ic ，SM ic Is equal to 1
Its meaning is a potential solution to characterize the objective function. Its initial position is determined by the following equation:
SM ij ＝SM minj +U(0,1)*(SM maxj -SM minj ) (13)
wherein U (0,1) produces a random number of 0-1 and SM maxj And SM minj Then the given upper and lower bounds of the individual jth dimension location information, respectively.
Then each individual needs to be optimized by learning the local leader position information of the group to which the individual belongs and a random identical group of individual position information:
SMnew ij ＝SM ij +U(0,1)*(LL kj -SM ij )+U(-1,1)*(SM rj -SM ij ) (14)
wherein LL kj Represents the individual SM i In the kth group, the j-dimension position information of the local leader; and SM rj Then represents any individual SM within the group k r The j-th dimension position information. And simultaneously, taking the optimal solution corresponding to the currently obtained optimal objective function value as the current global leader of the optimal solution. The position update formula of each individual at this time is as follows:
SMnew ij ＝SM ij +U(0,1)*(GL j -SM ij )+U(-1,1)*(SM rj -SM ij ) (15)
in the formula, GL j Dimension j position information representing current global leaderAnd (4) information. The individual fitness value characterizes the relative superiority of the individual in the current population, and is defined by the formula:
the location update probability of the individual at this time is:
and selecting the individual with the maximum fitness in the current population as a new global leader, and if the individual with the maximum fitness in the current population is the same as the new global leader, adding 1 to a global leader counter GlobalLeaderCount.
And selecting the individual with the maximum fitness in each group as a local leader in each group, and if the local leader is the same as the previous round of selection, adding 1 to a local leader counter.
If the local leader counter localreadercount of a certain group has not reached the given local leader count threshold localreaderlimit, the individuals within the group update their location information using the following equation:
SMnew ij ＝SM ij +U(0,1)*(GL j -SM ij )+U(0,1)*(SM ij -LL kj ) (18)
if the count value of the global leader counter globallerdercount reaches a given global leader count threshold globallerderlimit, the whole population needs to be split into more than one population. If the population number in the population reaches the preset maximum population number MG at the moment, all individuals are aggregated into a population, and the position of the global leader represents an approximate optimal solution at the moment, namely, the approximate optimal value of the formula (9) is obtained
And meanwhile, outputting the positions of the spider monkeys, namely the processing decision results corresponding to each power service.
The specific process involves: -
S31, first, according to the input data of the network topology N ═ U, C, R, E, the service importance level, the power service information attribute set a, and the basic compression ratio Δ, a time delay weighted and minimized objective function shown in formula (9) is constructed, and the specific steps of constructing the objective function include:
the edge server c may only process data of part of the services m, and the edge server c successively establishes two data packets 
And
to load the processed and unprocessed traffic data, respectively. Wherein the service data is not processed
Without waiting for processing delay
But can be directly encapsulated in a data packet
And sends the data to a system main station through the power Internet of things
Otherwise, the service data processed by the data processing
It will wait for the processing delay
And then the packet can be packed and sent to the system master station. According to the processing decision of the edge server c on each service m, the data volume of two data packets of the edge server c
And
respectively as follows:
wherein
And
and respectively loading the processed and unprocessed service data.
After the edge server c completes the encapsulation of the two data packets, the data packet obtained according to the specific routing rule is forwarded to the system master station d 0 The routing path of (1). Two data packets of the same edge server c are agreed in the invention
And
have the same forwarding routes. At the same time, if a certain section of link of network layer
Is simultaneously selected as the forwarding path of a plurality of edge servers c, then the link
Divide its bandwidth amount equally
To each edge server c. In order to characterize the routing relation of the data packet of the edge server, the invention uses binary variable
To characterize whether the data packet of the edge server c passes through the link 
The forwarding delay of the two data packets
And
are respectively defined as:
wherein the content of the first and second substances,
the data packet shares the link with the edge server when being forwarded in the network layer
The number of edge servers.
In summary, for the service m, the delay consists of two parts, namely forwarding delay and processing delay. Therefore, the service delay is defined as:
the optimization goal of the invention is to minimize the weighted delay sum of all services, where the delay weight coefficient of a service is the importance level k of each service m . In equation (8), the delay size of the traffic m on the edge server c
Depending on whether the data processing is done or not. Therefore, the objective function of the service delay optimization problem belongs to an integer programming problem, as shown in formula (9):
equation (9) shows whether the edge server c performs data processing on the service m or not under the condition that the objective of time delay weighted sum minimization is considered
The binary decision problem of (2), the corresponding constraints of the binary integer program are:
wherein constraints (10) - (12) are binary variable constraints of a binary integer program. Wherein, binary variables
And if the terminal u serves the service m, selecting the service m, wherein the value is 1, and otherwise, selecting the service m is 0. Binary variable 
To characterize whether the data packet of the edge agent c passes through the link
If passing through
The value is 1, otherwise 0. S32, an initial population of N individuals is generated, where N is determined according to the number of incoming power traffic. Wherein each individual is a C-dimensional vector, and the ith individual is recorded as SM ic ，SM ic Is equal to 1
Its meaning is a potential solution to characterize the objective function. The initial position is determined according to equation (13);
s33, each individual needs to update by learning the local leader location information of its belonging group and randomly selecting the same group of individual location information according to the probabilities of equations (16) - (17), and meanwhile, the current global leader is the optimal solution corresponding to the currently obtained optimal objective function value.
And S34, selecting the individual with the maximum fitness in the current population as a new global leader, and if the individual is the same as the previous round, adding 1 to a global leader counter GloballenderCount.
And S35, selecting the individual with the maximum fitness in each group as a local leader in each group, and if the local leader is the same as the previous round of selection, adding 1 to a local leader counter.
S36, if the local leader counter localreadercount of a certain group has not reached the given local leader count threshold localreaderlimit, the individuals in the group update their location information using equation (18):
S37, if the count value of the global leader counter globallerdercount reaches the given global leader count threshold globallerderlimit, the whole population needs to be split into one more population, that is, the step S33 is returned. Otherwise, all the individuals are aggregated into a group, and the algorithm execution is exited at S37, wherein the position of the global leader represents the approximate optimal solution, namely, the approximate optimal value of the acquisition formula (9)
Simultaneously outputs each spiderThe spider monkey position corresponds to the processing decision result of each power service.
Such as: one possible data processing decision scheme is the following table, with a final weighted delay sum of 892.69 ms.
In the corresponding edge server data processing decision scheme shown in the above table, the frequency of the edge server deciding to perform data processing on the four power services is 73.33%, 60% and 53.33%, respectively. The method of the invention effectively reduces the time delay in the edge service processing process and improves the efficiency and the performance of the network.
The invention discloses a differentiated service data flow processing decision method based on edge calculation in an electric power Internet of things. The change situation of the data volume processed by the edge server is quantified through an edge Internet of things proxy data processing model, and then the service experiment model is used for analyzing the time delay influence of data processing on the service. This changes the traffic weighted delay and minimization problem into a binary decision problem for the edge server to selectively handle traffic. This problem can be solved by using intelligent optimization algorithms. The invention can reduce the time delay in the edge service processing process and improve the efficiency and the performance of the network.
Claims (1)
1. The differentiated service data flow processing decision method based on edge calculation is characterized by comprising the following steps of:
s10, setting each service importance level accessed by the edge server: the method comprises the steps that the set of electric power services borne by an electric power communication network under the application scene of the method is set to be M, the importance level of the electric power service M belonging to the M is divided into 4 levels according to the time delay sensitivity degree, and the importance level of the service M is recorded as k m ，k m K is not more than K and K is 4;
s20, constructing a service processing depth model according to the service attributes and the network topology: selecting a service for data processing using an edge server cData volume S c Binary variable
Constructing a service processing depth model shown in a formula (3) by using a data base compression ratio delta and an information attribute set A; constructing a service processing depth model according to the service attributes and the network topology:
using binary variables respectively
And an information attribute set A to represent all information attributes contained in the service M and the | M | types of services served by the terminal u; therefore, the information attributes owned by the service m and the values of these information attributes are respectively represented by the binary information attribute set a m ＝(a 1 ,..,a i ,..a |A| ) And a set of attribute values B m ＝(b 1 ,..,b i ,..b |A| ) To characterize; in which the binary variable a i Data characterizing a service m includes an attribute a of 1 i And the value of the information attribute is b i ；
For each terminal u, the data information it transmits is characterized as a row vector
Since there are a plurality of terminals serving the same service in the edge server, the matrix for data information of the service m in the edge server c
To represent; matrix array
Has the dimension of
Wherein
Is edge server c coverageThe number of terminals of the internal service m; similarly, the matrix DM can be used when the edge server c receives the data information of all services c Is represented and matrix DM is c Has a dimension of
For any service m on the edge server c, the time delay comprises two parts, namely processing time delay and forwarding time delay; due to the computing power β of the edge server c c There is a limit, so all the traffic on the edge server c that selects to process data needs to wait for a processing delay
To avoid processing delay
If the size is too large, the edge server c selects to perform data processing on the data of only part of the services; processing latency of edge server c
The definition formula is:
wherein S is c And beta c Respectively calculating the service data volume and the server computing capacity; delta c And
respectively edge server c processing delay
And the two parameters are both the inherent property of the edge server and the size S of the received data c Irrelevant; basic processing delay
The time delay is a necessary time delay overhead of the initialization data processing process of the edge server c, and the time delay is determined by the hardware configuration of the edge server; since the edge server c selects to perform data processing only on part of the services, the edge server c selects the service data volume S for performing data processing c Characterised by the data matrix DM c I.e.:
wherein the content of the first and second substances,
the number of terminals serving the service m within the coverage of the edge server c; binary variable
Is a decision variable for the edge server c to determine whether to perform data processing on the service m;
the representation edge server c processes the data of the service m; thus um c Represents the total number of terminals served by the edge server c;
the ratio of the data size before and after the data processing is called the data processing depth phi of the edge server c c (ii) a However due to the set of information attributes a of the different services m m Composed of different information attributes, and therefore the data processing of different services on the edge server c
Is also different; defining the data processing depth phi of the edge server c c Comprises the following steps:
wherein, Delta is the data base compression ratio and is related to the effect of the software function used for processing the service;
s30, solving an optimal data processing decision scheme by applying an intelligent optimization algorithm, and setting a processing decision of each edge server for each service:
Because the edge server c only processes data of part of the service m, the edge server c can successively establish two data packets
And
to load processed and unprocessed service data respectively; wherein the service data is not processed
Without waiting for processing delay
But directly encapsulated in a data packet
And sends the data to a system main station through the power Internet of things
Otherwise, the service data processed by the data processing
It will wait for the processing delay
Then, the packets can be packaged and sent to a system main station; processing of individual services m according to edge server cDecision, data size of two data packets of edge server c
And
respectively as follows:
wherein
And
respectively loading the processed and unprocessed service data;
after the edge server c completes the encapsulation of the two data packets, the data packet obtained according to the routing rule is forwarded to the system master station d 0 The routing path of (a); two data packets for the same edge server c
And
have the same forwarding route; at the same time, if a certain section of link of network layer
Is simultaneously selected as the forwarding path of a plurality of edge servers c, then the link
Divide its bandwidth amount equally
To each edge server c; for characterizing routing relation of data packet of edge server, binary variable is used 
To characterize whether the data packet of the edge server c passes through the link
The forwarding delay of the two data packets
And
are respectively defined as:
wherein the content of the first and second substances,
the data packet shares the link with the edge server when being forwarded in the network layer
The number of edge servers;
in summary, for the service m, the delay consists of two parts, namely forwarding delay and processing delay; therefore, the service delay is defined as:
the optimization goal is to minimize the weighted delay sum of all services, where the delay weight coefficient of a service is the importance level k of each service m (ii) a In equation (8), the delay size of the traffic m on the edge server c
Depending on whether the data processing is done or not; therefore, the objective function of the service delay optimization problem belongs to an integer programming problem, and the delay weighting and minimizing problem is shown as formula (9):
equation (9) shows whether the edge server c performs data processing on the service m or not under the condition that the objective of time delay weighted sum minimization is considered
The binary decision problem of (2), the corresponding constraints of the binary integer program are:
wherein constraints (10) - (12) are binary variable constraints of a binary integer program; wherein, binary variables
If the terminal u serves the service m, the value of the service m is 1 when the service m is selected, otherwise, the value is 0; binary variable 
To characterize whether the data packet of the edge agent c passes through the link
If passing through
The value is 1, otherwise, the value is 0;
the intelligent optimization algorithm used for solving the service delay optimization problem is a spider monkey algorithm, and the steps are as follows:
firstly, generating an initial population consisting of N individuals, wherein N is determined according to the number of input power services; wherein each individual is a C-dimensional vector, and the ith individual is recorded as SM ic ，SM ic Is equal to 1
Its meaning is a potential solution to characterize the objective function; its initial position is determined by the following equation:
SM ij ＝SM minj +U(0,1)*(SM maxj -SM minj ) (13)
wherein U (0,1) produces a random number of 0-1 and SM maxj And SM minj Respectively setting upper and lower boundaries of the j-dimension position information of the individual;
then, each individual needs to perform a first step of location update optimization by learning the local leader location information of the group to which the individual belongs and a random same group of individual location information:
SMnew ij ＝SM ij +U(0,1)*(LL kj -SM ij )+U(-1,1)*(SM rj -SM ij ) (14)
wherein LL kj Represents the individual SM i In the kth group, the j-dimension position information of the local leader; and SM rj Then represents any individual SM within the group k r J-th dimension position information of (1); meanwhile, taking the optimal solution corresponding to the currently obtained optimal objective function value as the current global leader of the optimal solution; at this time, the position of each individual needs to be updated for the second time according to the global leader, and the formula is as follows:
SMnew ij ＝SM ij +U(0,1)*(GL j -SM ij )+U(-1,1)*(SM rj -SM ij ) (15)
In the formula, GL j J-th dimension position information representing a current global leader; the individual fitness value characterizes the relative superiority of the individual in the current population, and is defined by the formula:
wherein, f (x) i ) MinD for delay weighted sum minimization problem in objective function total (ii) a The location update probability of the individual at this time is:
wherein max _ fitness is the maximum value of all individual fitness values; selecting the individual with the maximum fitness in the current population as a new global leader, and if the individual with the maximum fitness in the current population is the same as the new global leader, adding 1 to a global leader counter GlobalLeaderCount;
selecting the individual with the maximum fitness in each group as a local leader in each group, and if the local leader is the same as the previous round of selection, adding 1 to a local leader counter;
if the local leader counter localreadercount of a certain group has not reached the given local leader count threshold localreaderlimit, the individuals within the group update their location information using the following equation:
SMnew ij ＝SM ij +U(0,1)*(GL j -SM ij )+U(0,1)*(SM ij -LL kj ) (18)
if the count value of the global leader counter GlobalLeaderCount reaches a given global leader count threshold GlobalLeaderLimit, the whole population needs to be split into one more population; if the population number in the population reaches the preset maximum population number MG at the moment, all individuals are aggregated into a population, and the position of the global leader represents an approximate optimal solution at the moment, namely, the approximate optimal value of the formula (9) is obtained 
Meanwhile, outputting the positions of the spider monkeys, namely the processing decision results corresponding to each power service;
the specific process involves:
s31, first, according to the input data of the network topology N ═ U, C, R, E, the service importance level, the power service information attribute set a, and the basic compression ratio Δ, a time delay weighted and minimized objective function shown in formula (9) is constructed, and the specific steps of constructing the objective function include:
the edge server c may only process data of part of the services m, and the edge server c successively establishes two data packets
And
to load processed and unprocessed service data respectively; wherein the service data is not processed
Without waiting for processing delay
But can be directly encapsulated in a data packet
And sends the data to a system main station through the power Internet of things
Otherwise, the service data processed by the data processing
It will wait for the processing delay
Then, the packets can be packaged and sent to a system main station; according to the processing decision of the edge server c on each service m, the data volume of two data packets of the edge server c
And
respectively as follows:
wherein
And
respectively loading the processed and unprocessed service data;
after the edge server c completes the encapsulation of the two data packets, the data needs to be obtained according to a specific routing rule Packet forwarding to system Master d 0 The routing path of (a); two data packets for the same edge server c
And
have the same forwarding route; at the same time, if a certain section of link of network layer
Is simultaneously selected as the forwarding path of a plurality of edge servers c, then the link
Divide its bandwidth amount equally
To each edge server c; for characterizing routing relation of data packet of edge server, binary variable is used
To characterize whether the data packet of the edge server c passes through the link
The forwarding delay of the two data packets
And
are respectively defined as:
wherein the content of the first and second substances,
the data packet shares the link with the edge server when being forwarded in the network layer
The number of edge servers;
in summary, for the service m, the delay consists of two parts, namely forwarding delay and processing delay; therefore, the service delay is defined as:
the optimization goal is to minimize the weighted delay sum of all services, where the delay weight coefficient of a service is the importance level k of each service m (ii) a In equation (8), the delay size of the traffic m on the edge server c
Depending on whether the data processing is done or not; therefore, the objective function of the service delay optimization problem belongs to an integer programming problem, as shown in formula (9):
equation (9) shows whether the edge server c performs data processing on the service m or not under the condition that the objective of time delay weighted sum minimization is considered 
The binary decision problem of (2), the corresponding constraints of the binary integer program are:
wherein constraints (10) - (12) are binary variable constraints of a binary integer program; wherein, binary variables
If the service m is served for the terminal u, the value is 1 when the service m is selected, otherwise, the value is 0; binary variable
To characterize whether the data packet of the edge agent c passes through the link
If passing through
The value is 1, otherwise, the value is 0;
s32, generating an initial population consisting of N individuals, wherein N is determined according to the number of input power services; wherein each individual is a C-dimensional vector, and the ith individual is recorded as SM ic ，SM ic Is equal to 1
Its meaning is a potential solution to characterize the objective function; the initial position is determined according to equation (13);
s33, then, each individual needs to learn the local leader position information of the group to which the individual belongs, randomly select the same group of individual position information according to the probabilities of the formulas (16) - (17) to update, and meanwhile, the optimal solution corresponding to the currently obtained optimal objective function value is used as the current global leader of the individual;
s34, selecting the individual with the maximum fitness in the current population as a new global leader, and if the individual is the same as the previous round, adding 1 to a global leader counter GloballenderCount;
S35, selecting the individual with the maximum fitness in each group as a local leader in each group, and if the selection is the same as the selection in the previous round, adding 1 to a local leader counter;
s36, if the local leader counter localreadercount of a certain group has not reached the given local leader count threshold localreaderlimit, the individuals in the group update their location information using equation (18):
s37, if the count value of the global leader counter globallerdercount reaches the given global leader count threshold globallerderlimit, the whole population needs to be split into one more population, that is, the step S33 is returned. Otherwise, all the individuals are aggregated into a group, and the algorithm execution is exited at S37, wherein the position of the global leader represents the approximate optimal solution, namely, the approximate optimal value of the acquisition formula (9)
And meanwhile, outputting the positions of the spider monkeys, namely the processing decision results corresponding to each power service.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111002096.1A CN113676357B (en) | 2021-08-30 | 2021-08-30 | Decision method for edge data processing in power internet of things and application thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111002096.1A CN113676357B (en) | 2021-08-30 | 2021-08-30 | Decision method for edge data processing in power internet of things and application thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113676357A CN113676357A (en) | 2021-11-19 |
| CN113676357B true CN113676357B (en) | 2022-07-29 |
Family
ID=78547364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111002096.1A Active CN113676357B (en) | 2021-08-30 | 2021-08-30 | Decision method for edge data processing in power internet of things and application thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113676357B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114390055B (en) * | 2021-12-08 | 2023-12-05 | 广东电网有限责任公司 | Cloud edge cooperative data acquisition method and system suitable for power industry |
| CN116701962B (en) * | 2023-08-07 | 2023-10-27 | 北京电科智芯科技有限公司 | Edge data processing method, device, computing equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112702401A (en) * | 2020-12-15 | 2021-04-23 | 北京邮电大学 | Multi-task cooperative allocation method and device for power Internet of things |
| CN112752302A (en) * | 2021-01-05 | 2021-05-04 | 全球能源互联网研究院有限公司 | Power service time delay optimization method and system based on edge calculation |
| CN113011678A (en) * | 2021-04-08 | 2021-06-22 | 国网辽宁省电力有限公司信息通信分公司 | Virtual operation platform operation control method based on edge calculation |
| WO2021139537A1 (en) * | 2020-01-08 | 2021-07-15 | 上海交通大学 | Power control and resource allocation based task offloading method in industrial internet of things |
| CN113268341A (en) * | 2021-04-30 | 2021-08-17 | 国网河北省电力有限公司信息通信分公司 | Distribution method, device, equipment and storage medium of power grid edge calculation task |
| CN113286269A (en) * | 2020-10-31 | 2021-08-20 | 国网河南省电力公司经济技术研究院 | Electric power communication network resource management method based on edge calculation |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109818865B (en) * | 2019-03-11 | 2020-09-18 | 江苏君英天达人工智能研究院有限公司 | SDN enhanced path boxing device and method |
| CN110717302B (en) * | 2019-09-27 | 2022-07-01 | 云南电网有限责任公司 | Edge computing terminal equipment layout method for real-time online monitoring service of power grid |
| CN112187534B (en) * | 2020-09-21 | 2021-09-24 | 上海交通大学 | Task unloading method based on multi-hop transmission in industrial Internet of things |
-
2021
- 2021-08-30 CN CN202111002096.1A patent/CN113676357B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021139537A1 (en) * | 2020-01-08 | 2021-07-15 | 上海交通大学 | Power control and resource allocation based task offloading method in industrial internet of things |
| CN113286269A (en) * | 2020-10-31 | 2021-08-20 | 国网河南省电力公司经济技术研究院 | Electric power communication network resource management method based on edge calculation |
| CN112702401A (en) * | 2020-12-15 | 2021-04-23 | 北京邮电大学 | Multi-task cooperative allocation method and device for power Internet of things |
| CN112752302A (en) * | 2021-01-05 | 2021-05-04 | 全球能源互联网研究院有限公司 | Power service time delay optimization method and system based on edge calculation |
| CN113011678A (en) * | 2021-04-08 | 2021-06-22 | 国网辽宁省电力有限公司信息通信分公司 | Virtual operation platform operation control method based on edge calculation |
| CN113268341A (en) * | 2021-04-30 | 2021-08-17 | 国网河北省电力有限公司信息通信分公司 | Distribution method, device, equipment and storage medium of power grid edge calculation task |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113676357A (en) | 2021-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yu et al. | Toward resource-efficient federated learning in mobile edge computing | |
| CN111447083B (en) | Federal learning framework under dynamic bandwidth and unreliable network and compression algorithm thereof | |
| WO2021169577A1 (en) | Wireless service traffic prediction method based on weighted federated learning | |
| Yu et al. | An intelligent game-based offloading scheme for maximizing benefits of IoT-edge-cloud ecosystems | |
| CN113676357B (en) | Decision method for edge data processing in power internet of things and application thereof | |
| CN110968426B (en) | Edge cloud collaborative k-means clustering model optimization method based on online learning | |
| CN110493360A (en) | The mobile edge calculations discharging method of system energy consumption is reduced under multiserver | |
| CN111475274A (en) | Cloud collaborative multi-task scheduling method and device | |
| WO2023040022A1 (en) | Computing and network collaboration-based distributed computation offloading method in random network | |
| CN113989583A (en) | Method and system for detecting malicious traffic of internet | |
| CN110290077B (en) | Industrial SDN resource allocation method based on real-time service configuration | |
| CN114374605A (en) | Dynamic adjustment and migration method for service function chain in network slice scene | |
| CN114885388B (en) | Multi-service type self-adaptive switching judgment method combining RSS prediction | |
| Li et al. | DQN-enabled content caching and quantum ant colony-based computation offloading in MEC | |
| CN114615744A (en) | Knowledge migration reinforcement learning network slice general-purpose sensing calculation resource collaborative optimization method | |
| Wu et al. | Link congestion prediction using machine learning for software-defined-network data plane | |
| Cui et al. | Multi-Agent Reinforcement Learning Based Cooperative Multitype Task Offloading Strategy for Internet of Vehicles in B5G/6G Network | |
| CN115002031B (en) | Federal learning network flow classification model training method, model and classification method based on unbalanced data distribution | |
| CN114615705B (en) | Single-user resource allocation strategy method based on 5G network | |
| CN111723978B (en) | Index evaluation method for adapting to various power business difference demands based on virtual mapping | |
| CN116109058A (en) | Substation inspection management method and device based on deep reinforcement learning | |
| CN114785692A (en) | Virtual power plant aggregation regulation and control communication network flow balancing method and device | |
| CN115361453A (en) | Load fair unloading and transferring method for edge service network | |
| Hoiles et al. | Risk-averse caching policies for YouTube content in femtocell networks using density forecasting | |
| CN112969157A (en) | Network load balancing method for unmanned aerial vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 010041 No. 11, Ordos East Street, Ruyi Development Zone, Saihan District, Hohhot City, Inner Mongolia Autonomous Region Applicant after: STATE GRID EAST INNER MONGOLIA ELECTRIC POWER Co.,Ltd. Applicant after: North China Electric Power University Address before: 102206 No. 2 Nong Road, Zhu Xin Zhuang, Beijing, Changping District Applicant before: NORTH CHINA ELECTRIC POWER University Applicant before: STATE GRID EAST INNER MONGOLIA ELECTRIC POWER Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |